This question comes up every month. And every month, people who have no idea how the model works post inaccurate analysis.
Birth-Death Adjustment Yet Again
The above statement by Peter Schiff and similar statements by others are false. Schiff should know better. Even prominent economists make the same mistake.
Unfortunately, others repeat this stuff over and over without anyone doing any basic research that immediately disproves the claim.
I will go over the details of why Schiff is wrong, but first let’s discuss the process. Here is a link to a BLS discussion of the Birth-Death Adjustment model.
The birth-death model pertains to the birth and death of corporations not individuals. When corporations go into business they add workers and those workers are not in the BLS sample. So the BLS uses these birth-death adjustments to tweak its monthly jobs reports.
If this seems totally convoluted, it’s because it is totally convoluted.
Instead of relying on tax data, the BLS relies on the model. The BLS rationale is that not everyone reports tax data in the reference week, once a month, that determines the report.
The reference week and the relatively small sample vs quarterly reports are sources of error. In general, people do not look at the quarterly reports because they prefer the instant gratification of highly suspect numbers.
I discuss quarterly numbers below, but first let’s continue with other sources of error in the monthly reports.
BLS Reference Week
The BLS sample is about 6.5 percent of the data, taken once a month ion the reference week.
Like Peter Schiff, Danielle DiMartino Booth, and Lacy Hunt, I am highly skeptical of this jobs report.
Every month in my Jobs Report Analysis, I post these statements “For those who follow the Birth-Death numbers, I retain this caution: Do not subtract the reported Birth-Death number from the reported headline number. That approach is statistically invalid.“
But that is what Schiff did. It’s not just Schiff. Sadly, prominent economists make the same mistake.
Birth-Death Adjustment
For convenience, here is the lead chart again.
The BLS birth-death adjustment added 231,000 jobs to the unadjusted total in May. It added 603,000 to the unadjusted numbers for the year.
Mass layoffs have nothing to do with this, nor do bankruptcies. It’s an estimate of business openings and closings. Were there 603,000 new jobs this year created by new businesses? 231,000 in May?
Color me very skeptical, but I have no idea what the number is. Nor does anyone else. In general, more businesses open than close over time, but the estimates are highly likely to be very wrong at turns.
Seasonal Adjustments
Importantly, the birth-death numbers are not seasonally adjusted.
You should not subtract these unadjusted numbers from the reported seasonally adjusted jobs numbers.
My Conversation With the BLS
- Mish: How much did the Birth-Death adjustment impact the seasonally-adjusted gain?
- BLS: Because of our methodology, we don’t know.
- Mish: What is the seasonally-adjusted Birth-Death number?
- BLS: Because of our methodology, we don’t know.
That conversation was not from last month but rather from over 10 years ago. That’s a synopsis. The BLS was polite, not abrupt as presented above.
Nothing has changed except tweaks to the model that supposedly improve it, and likely do (but that does not imply the model is accurate at economic turns).
The following process, derived from a long conversation with the BLS, explains why the BLS itself does not know how the birth-death number impacts the headline number.
Five-Step Process
- The BLS calculates the unadjusted birth-death number. It’s an undisclosed magic hat based on experience.
- The birth-death adjustment is made to unadjusted overall estimate for the month.
- The BLS takes the unadjusted total, including the adjustment, then seasonally adjusts the number.
- The BLS then compares the seasonally-adjusted number for this month and subtracts the seasonally-adjusted number from last month.
- The difference between the seasonally adjusted numbers is the headline total.
That’s why the BLS cannot say, nor can anyone else say, how much the birth death model impacted the seasonally-adjusted headline number.
The following chart should make it clear why one cannot subtract unadjusted numbers from seasonally-adjusted ones.
Nonfarm Payrolls SA vs NSA
I created the above chart today to show the silliness of typical birth-death subtraction analysis.
The actual math shows it’s even worse than it looks.
Actual Math for May 2024
- The unadjusted total payroll for May including the B/D adjustment is 158,918,000 (158.9 million)
- Of that unadjusted total, 231,000 (231 thousand) were from the birth death adjustment
- The BLS then seasonally adjusted that number to arrive at the reported SA headline number of 158,543,000.
- Ta Da! The reported headline number for May (158,543,000) is 375,000 lower than the unadjusted report total (158,918,000).
Question of the Day
Q: How much did the birth death adjustment boost the reported May report of 158,543,000?
A: You don’t know, I don’t know, and amazingly not even the BLS can tell you because the process is so convoluted.
Whatever the number, except by extreme accident, it certainly is not 231,000 and probably not anything close.
Bonus Question of the Day
Q: Is it possible this Birth-Death procedure is artificially padding payrolls?
A: Of course. But you don’t know how much, I don’t know how much, and not even the BLS can tell you!
However, we can say that sensible people do not believe this jobs report. Let’s put a spotlight on why.
QCEW Discrepancy
With little doubt, the QCEW data tells us something, perhaps many things including the birth-death adjustment, is artificially padding the monthly numbers.
QCEW (Quarterly Census of Employment and Wages) is a count of Unemployment Insurance (UI) administrative records submitted by 11.9 million establishments. That’s about 99 percent of the data.
Nonfarm Payrolls are are from the Establishment Survey (CES). The sample survey was 666,000 individual worksites in 2023. That’s about 5.6 percent of the data.
Clearly, QCEW will be far more accurate than these monthly convolutions.
How Much Does the BLS Overstate Monthly Jobs?
I discussed the above question in my June 6, 2024 post How Much Does the BLS Overstate Monthly Jobs? Here’s the Answer
Understanding the Discrepancies
If the difference was nearly constant, say ~2.5 million, it would not matter. But if sharply increasing or decreasing numbers happen near recessions, the discrepancies do matter.
Comparing NSA numbers, Nonfarm payrolls are 3.42 million higher than QCEW numbers. That is the largest difference since 3.46 million at the height of the Covid pandemic.
The only other 3+ million difference was in the third quarter of 2020.
After discussing this with Lacy Hunt, he suggested that I look at percentages. Here are two new charts that better show what’s happening.
Y-O-Y Percent Percentage Change in Nonfarm Payrolls vs QCEW
The percentages look tiny. But they are trending, not random. And multiplied by 150 million jobs, the numbers are significant.
I recall reading a BLS statement that anything over 0.1 percent is significant, but I cannot locate that comment.
Let’s show the result in numbers instead of percentages.
Year-Over-Year Change in Nonfarm Payrolls vs QCEW in Thousands
Note that coming out of recessions, the BLS monthly reports understate jobs growth. Heading into recessions, when data is weakening, the opposite occurs.
Another Bizarro Jobs Report – Payrolls Rise 272,000 Employment Drop 408,000
For more details on the May jobs report, please see Another Bizarro Jobs Report – Payrolls Rise 272,000 Employment Drop 408,000
Once again, rational people wonder what’s going on with the jobs report. The discrepancy between the trend in jobs and employment surges again.
The allegedly stunning jobs report shows that over the past year jobs are up 2.76 million from a year ago but employment is up only 376,000.
The report also shows full-time Employment is -1,163,000 vs a year ago.
Click on the preceding link for more details.
A Second-Quarter Recession This Year Looks Increasingly Likely
I believe we are headed for recession this year. A recession may have even started.
For discussion, please see A Second-Quarter Recession This Year Looks Increasingly Likely
The GDPNow model I refer to in the above post is now sharply higher, some of it based on the May jobs report.
Cumulative Birth-Death Errors
Given that I put a spotlight on the significance of 0.1 percent errors, what if the Birth-Death numbers accumulate significant errors?
Here is the math for May.
Birth-Death 231,000 / (158,918,000 – 231,000) = 0.0014556
Call it 0.15 percent.
A cumulative monthly error like this could be a key source of the QCEW errors at economic turns, although 0.015 percent per month seems ridiculously high.
For the sake of argument, let’s assume the Birth-Death model is on the high side by ~0.10 to ~0.15 percent per quarter (not per month).
If the Birth-Death error wrongly accumulates every month in the same direction at economic turns (say 0.10 to 0.15 per quarter), that is a big piece of the discrepancy puzzle.
Conclusion
The birth-death model may be a significant source of the QCEW vs CES discrepancy. But the whole process is so convoluted, in so many places, that it is hard to pinpoint a precise source of the errors.
The Birth-Death adjustment could be a major or minor source of the error.
Regardless, we can say that QCEW and other weakening hard data provide ample reason to mock the current report.
Through the fourth quarter of 2023, QCEW suggest the monthly CES report overstates jobs by about 735,000 jobs, all in the second half of 2033. That’s 125,500 jobs a month, on average.
What is the discrepancy now for 2024 Q1?
The current CES to QCEW discrepancy is 0.45 percent. Will the discrepancy for 2024-Q1 be in the range (0.055-0.075)? I will be watching this.
Schiff is right to be suspicious, and so is everyone else. But let’s not use invalid birth-death monthly math to come up with a number for this month, or any month.
Excellent utube video by Daniella Martino Booth. Link: link to youtube.com
Excellent discussion by Mish regarding the convoluted stats by BLS as to employment and drilling down how the BLS arrives at their modeling numbers.
You can slice and dice it however. These numbers are not recessionary. Is the economy slowing? Yes. Until mass layoffs go way up and openings go way down, there won’t be a recession.
It has been over a year of conflicting data. I’ve been in a constant battle between deflation, entrenched inflation, and somewhere in the soft landing middle. For example, look at the six month trends in inflation – not looking good for the side Mish seems to take. Really, inflation troughed a full year back now – CPI was 3.0 in June, 2023.
The main takeaway I’m getting from comments like yours and the market by large – talk that high rates are causing inflation, even Fed members questioning if restrictive rates are doing anything at all… – is that we’ve been in the hiking cycle for so long now that we’ve all seemingly forgotten about long and variable lags. High rates will slow the economy, there was just so much excess liquidity pumped in that it’s taking a very long time to drain (+Biden and camp RhinoDino fighting the Fed). Plus, QT is still countering money growth. It’s a matter of time, and the Fed is usually late to the party which is probably why folks like Mish are looking for cuts sooner rather than later.
One more point scored for the economy is tanking side of the debate. New York tax receipts are in the toilet. Down 5% or something compared to last year. California tax receipts have been well below expectations all year, but have risen from -7ish % to 2.7%. Their tax collections still well up, and the tax filing deadline changes are skewing the comps, but still they expected better. California unemployment is up pretty strong. They usually lead us into recession.
The May US BLS Disability numbers are in.
The US population 16 & Over has been rising again every month since January and sits less than 100k from new all time high.
The Civilian Labor Force 16 & Over went to a slight new all time high.
This trend is going in the wrong direction 4 years after the pandemic & furthermore the US civilian labor force which should be healthier than the general US population is making a new high.
Again why is is this not an interesting phenomenon to the watchdog institutions?
link to t.me
What could be disabling so many????
Statistics, that item where by literally trillions of currency unit are placed for investment/speculation in markets.
Statistics that no one not even the best and brightest can have any confidence that numbers as they are presented reflect underlying fundamentals as to economic conditions.
This does not paint a very confidence building picture of where people put their life savings and retirement funds looking for some return whether that be some percent gain or even a return of Capital.
If it is by design or incompetence turns out that much if not most of financial markets are based upon inaccurate and likely Fraudulent statistics.
Since markets have always been built upon some form of Trust in Fiduciary responsibility, what happens as the curtain gets pulled back and people learn that the man behind said curtain is a fraud?
This has all happened before and ensuing crackups have lead to ruin and chaos.
Just saying. Is that Fed Put really big enough to count upon now that the consequences of deploying it are manifesting? Would suggest Fed Put will not be able to save everyone.
The employment-to-population ratio is not only below the pre-pandemic rate but way below the pre-pandemic trend; in fact, the May reading was the same rate as Mar ’21 and indicates millions are missing from the labor market link to t.me
Hmmm…. wonder what caused that?
As The Gateway Pundit reported on Thursday, the Kansas City Chiefs halted all team activities after defensive lineman BJ Thompson suffered a seizure and soon after went into cardiac arrest during a special teams meeting.
In a statement on Thursday night, Thompson’s agent Chris Turnage said, “BJ is still unconscious at this point, but he’s stable, and his vitals are good.”
“His family asks for your continued prayers,” added Turnage.
– Mass layoffs have nothing to do with this, nor do bankruptcies. It’s an estimate of business openings and closings. Were there 603,000 new jobs this year created by new businesses? 231,000 in May?
Color me very skeptical, but I have no idea what the number is. Nor does anyone else. In general, more businesses open than close over time, but the estimates are highly likely to be very wrong at turns.
> I will say it once again, as it deserves to be said, until everyone realizes it’s all made up as they go. They change the data all the time, and to a point where it makes no sense. Nothing new here folks, as this has been going on for about 3 years now perhaps…
Most data from the Government is heavily skewed to meet whatever their agenda is at the time. The data matches up with what they want to show for numbers, to get something passed, delayed, or removed.
Possible examples would be:
1. The Afghanistan Withdrawal, where massive amounts of weapons were left behind, and that was obviously to force much more money into the Military for weapons advancement, and replacements.
2. The Ukrainian Weapons/Money Giveaway, were massive amounts of. Old Useless Weapons were shipped to Ukraine to be used as fodder, and to force much more money into the Military for weapons advancement, and replacements. They also took $Billions$ of our Citizens tax dollars and paid off everyone involved with the underground experimental goings on (think Vaccines/Illnesses), that had to be destroyed and people paid to keep quiet. Oh, and some nice Pensions being paid to these NGO’s and the like. Wish Americans got Pensions like the Ukrainian Workers (*wink,wink”).
3. Fauci and Company, need I say anything else there?
4. Wuhan Labs, need I say anything else there?
5. Etc.
It’s all really such a Very Sad Joke being played on Americans right now, and has been for a number of years now… Bring on November!!!
I remember Alan Greespan said years ago that he didn’t have enough good data to make informed decisions.
But he continued in the game instead closing the shop, with that marvelous results we all know about.
Solid stuff, as usual.
Mish, have you seen Anna Wong’s (Bloomberg) tweet about Homebase data derived estimates diverging from BLS numbers?
I’m surprised no one has designed a tool to scrape job openings at major companies. Walmart alone has 43,000 full time positions open right now. That’s just one employer.
link to careers.walmart.com
Amazon has tons of jobs open too but I couldn’t find a total anywhere, it’s city by city and there are hundreds of jobs in each city.
Let’s not forget that at least 2 million boomers will leave the labor force this year. Social security snapshot should be out next week, we’ll see what it shows.
Walmart is existing. Those numbers are in the monthly samples.
The REAL labor force is much bigger than reported. The largest increases are in leisure and hospitality, construction and in the unreported online businesses bs. In the last 20 years about 15/20 million entered the US. Many are in the black market. Some are in street corners. They must be added. Many don’t work even one hour/week. Many are under the poverty line. The Real income is lower. The Real unemployment is much higher, well above 5%/6%.
Mish, I read almost every article you post. The labor reports, due to their supposed importance to the Fed are the most frustrating.
Are there any labor reports that you believe have any credence?
Any method to simplify tracking the labor reports that can be more useful?
The QCEW report based on 99% of the data should be very accurate.
But it’s June.
The Latest QCEW report is for December 2023.
There is little else to go by.
This Report?
“QCEW Revisions” Tables and Data Now Available
“QCEW revisions” tables and data for establishments, employment, and wages now available for recent years via link to bls.gov.
U.S. Bureau of Labor Statistics Logo
U.S. BUREAU OF LABOR STATISTICS
OEUS/DASLT, PSB Suite 4860
2 Massachusetts Avenue NE
Washington, DC 20212-0001
Telephone:1-202-691-6567
link to bls.gov
Contact QCEW
RESOURCES:
Inspector General (OIG)
Budget and Performance
Department of Labor Grants
No Fear Act
USA.gov
Vote.gov
Are you kidding me? That’s the resources that you are trusting to be 99% Accurate? C’mon Mish, I know You know way better than that!!!
It’s 99 percent of the data not 99% accurate.
But there are no better stats.
The problem is it lags a minimum of 5 months and is subject to revisions.
But if you don’t have any faith in that, then you should never believe anything ever.
No stats ever are better than that
“My bad” on the % mix up, sort of irrelevant, but the report is simply ridiculous on its own merits that it provides. They go through literally 6 iterations of Revisions. This means that the data is absolutely useless, by the time it’s corrected to its actual input value. Your comment Mish “ No stats ever are better than that” is ludicrous on its own merit as well, If your standards are “5 Month Minimum Lags”, and Multiple iterations of Revisions, rendering the stats absolutely useless for any concrete decisions to be made in time, and far too many variables to obtain any sort of useable cherry picked and/or hunches or otherwise. Any decision made, is simply not based on reliable information, and therefor unusable for anything of value. So therefor I have very little faith in those numbers as usable to any paying client, as a result.
The Resources are also strange, for the purpose as well IMO. “No Fear Act” – The No Fear Act governs the process of reimbursements to the Treasury Department’s judgment fund by agencies, from their budgets, for judgments against agencies and settlements for discrimination in the workplace. The Act requires Federal agencies to be accountable for violations of antidiscrimination and whistleblower protection laws, in part by requiring that each Federal agency post quarterly on its public Web site, certain statistical data relating to Federal sector equal employment opportunity complaints filed with each agency.
What exactly does that have to do with this report and/or the numbers? Others as well are listed, with no reasoning I can come up with, to be a resource for this report. Am I missing something? I just find this report strange, useless, and full of constant stirs overtime?
Good Luck with it…
Those in the black market might be in the birth/death report, but are not in QCEW.
I’m so confused. It reassuring but it isn’t that the BLS has no clue. There has to be a better way. Time to scrap the current system and start over.
Any wonder the ATL FED’s GDP NOW is so often overstated? They take (mostly) government data that is seasonalized , revised and stylized and use it to come up with GDP estimates. Good for them for trying but it’s a complete waste of time given the unreliable data they must rely on.
GDPNow has been amazingly accurate for a couple of years. Kudos to Pat Higgins, the inventor.
He overstated GDP 20 straight times in the latest quarter. By a wide margin. It’s incredibly inaccurate and pointless.
In the real world a guy like this would be fired.
If the raw numbers are birth/death adjusted, then seasonally adjusted, but the birth/death numbers are not understood and are thought bogus, then to what extent must we consider the official seasonally adjusted numbers to be bogus?
I do not have long term analysis. Data is there, but historical monthly QCEW data is not easy to work with.
Note that the discrepancy from 2022 Q1 Through 2023 Q2 averaged -0.1.
Huge miss in last 2 quarters and coming out of the Covid recession.
“I am highly skeptical of this jobs report.”
Um, come to think about it, are there ANY government or Fed numbers reported with which we can have a reasonable level of confidence? Candidly, can you name one?
QCEW quarterly
As Dr. William Barnett of Divisia Aggregates says: “The Fed should establish a “Bureau of Financial Statistics”.
How much faith can be placed upon ANY government statistic?
For Government Workers 100%
For Everyone else 0%
I go with the theory, that if most data is incorrect, skewed, manipulated etc, then it’s a damn good chance ALL Data is incorrect, skewed, manipulated etc. unless of course, it happens to match up with what’s truly going on. Sort of the blind squirrel theory…